What Half a Million Tech Layoffs Look Like in the Data

January 2023 is still the worst single month of the correction. Google, Microsoft, Amazon and Salesforce announced about 48,000 cuts within twenty days of each other.

We pulled together the public trackers (layoffs.fyi, the Challenger Gray reports, company filings) and our own curated event set to get the shape of the 2022 to 2025 layoffs. Numbers first.

A peak in 2023, then a second wind

Tech layoffs came to roughly 165,000 in 2022, peaked at about 264,000 in 2023 and fell to about 152,000 in 2024. Then they picked up again. October 2025 was the worst October for announced US job cuts, across all industries, in more than two decades.

For scale, the correction removed about as many people as US big tech added in 18 months of pandemic hiring. The industry never shrank below its 2019 size. It gave back the bubble, then kept trimming for other reasons.

Recruiters were hit hardest, security barely at all

Function-level analyses (LinkedIn's Economic Graph, revelio-style workforce data) keep ranking recruiting and HR as hardest hit for their size. The function that grows with hiring had little to do once hiring stopped. Customer support, marketing and program or project management come next. Engineering was cut too, but consistently below its share of headcount. Security was barely touched.

The biggest hiring-side hit landed on new grads, whose hiring is down about 50% from 2019.

How much of it was AI

In our event set, cuts with a credible AI link covered essentially 0% of affected people in 2022. In 2023 it was single digits (IBM's pause, Chegg, Stack Overflow). In 2024 it was roughly a fifth, with SAP, Intuit, Cisco, Dropbox and Duolingo. By 2025 it was approaching half of the people we tracked, across Amazon, Microsoft, Salesforce, Accenture, TCS, HP, Workday and CrowdStrike.

Two caveats. Some of the attribution is fashion (see our piece on AI layoffs and corporate spin). And the biggest AI effect, the jobs that were never created, shows up in no layoff dataset.

That second caveat matters most, because some of what hurts the market never produces a countable number. Teams that would have added ten people add six. Companies follow the IBM model of attrition absorption, where nobody gets fired, departures just don't get replaced, and AI picks up the work. And fewer people are expected to deliver the same output, with AI assumed to fill the gap.

None of that makes a headline, and all of it tightens the market. If job hunting feels worse than the layoff totals suggest, you're reading it right.